Hostname: page-component-745bb68f8f-kw2vx Total loading time: 0 Render date: 2025-02-05T20:10:34.487Z Has data issue: false hasContentIssue false
  • English
  • Français

ON THE LINEAR COMBINATION OF LAPLACE RANDOM VARIABLES

Published online by Cambridge University Press:  31 August 2005

Saralees Nadarajah  and
Samuel Kotz
Saralees Nadarajah
Affiliation:
Department of Statistics, University of Nebraska, Lincoln, NE 68583, E-mail: snadaraj@unlserve.unl.edu
Samuel Kotz
Affiliation:
Department of Engineering Management and Systems Engineering, George Washington University, Washington, DC 20052, E-mail: kotz@gwu.edu
Rights & Permissions [Opens in a new window]

Abstract

The distribution of the linear combination αX + βY is derived when X and Y are independent Laplace random variables. Extensive tabulations of the associated percentage points are also given. The work is motivated by examples in automation, control, fuzzy sets, neurocomputing, and other areas of informational sciences.

Type
Research Article
Information
Probability in the Engineering and Informational Sciences , Volume 19 , Issue 4 , October 2005 , pp. 463 - 470
Copyright
© 2005 Cambridge University Press

1. INTRODUCTION

For given random variables X and Y, the distribution of linear combinations of the form αX + βY is of interest in problems in automation, control, fuzzy sets, neurocomputing, and other areas of informational sciences. The distribution of αX + βY has been studied by several authors, especially when X and Y are independent random variables and come from the same family. For instance, see Fisher [8] and Chapman [3] for the Student's t family, Christopeit and Helmes [4] for the normal family, Davies [5] and Farebrother [7] for the chi-squared family, Ali [2] for the exponential family, Moschopoulos [12] and Provost [15] for the gamma family, Dobson, Kulasmaa, and Scherer [6] for the Poisson family, Pham-Gia and Turkkan [14] and Pham and Turkkan [13] for the beta family, Kamgar-Parsi, Kamgar-Parsi, and Brosh [11] and Albert [1] for the uniform family, Hitezenko [9] and Hu and Lin [10] for the Rayleigh family, and Witkovský [17] for the inverted gamma family.

In this article, we study the distribution of αX + βY when X and Y are independent Laplace random variables with probability density functions (p.d.f.s)

and

respectively, for −∞ < x < ∞, −∞ < y < ∞, λ > 0, μ > 0, −∞ < θ < ∞, and −∞ < φ < ∞. We assume without loss of generality that α > 0. Extensive tabulations of the associated percentage points are also provided.

2. CDF

Theorem 1 derives explicit expressions for the cumulative distribution function (c.d.f.) of αX + βY in terms of elementary functions.

Theorem 1: Suppose X and Y are distributed according to (1) and (2), respectively. The c.d.f. of Z = αX + βY can be expressed as one of the following:

1. If β < 0 and φ < (z − θα)/β, then

2. If β < 0 and φ > (z − θα)/β, then

3. If β > 0 and φ < (z − θα)/β, then

4. If β > 0 and φ > (z − θα)/β, then

For (3)–(6), p = λ(z/α − θ) − λβφ/α and r = μφ − μ(z − θα)/β.

Proof: The c.d.f. F(z) = Pr(αX + βY ≤ z) can be expressed as

where F(·) inside the integral denotes the c.d.f. corresponding to (1) and given by

The results in (3)–(6) follow by elementary integration of (7). █

The following corollaries provide the c.d.f.s for the sum and the difference of Laplace random variables.

Corollary 1: Suppose X and Y are distributed according to (1) and (2), respectively. Then the c.d.f. of Z = X + Y can be expressed as one of the following:

1. If φ < z − θ, then

2. If φ > z − θ, then

In the above, p = λ(z − θ − φ) and r = μ(θ + φ − z).

Proof: Set α = 1 and β = 1 into (5) and (6). █

Corollary 2: Suppose X and Y are distributed according to (1) and (2), respectively. Then the c.d.f. of Z = X − Y can be expressed as one of the following:

1. If φ < θ − z, then

2. If φ > θ − z, then

In the above, p = λ(z + φ − θ) and r = μ(z + φ − θ).

Proof: Set α = 1 and β = −1 into (3) and (4). █

3. PERCENTILES

In this section, we provide tabulations of percentage points zp associated with the c.d.f. of Z = αX + βY. These values are obtained by numerically solving the equation F(zp) = p, where F is given by one of the eight formulas in the previous section. The computation involved is elementary and was performed using the algebraic manipulation package MAPLE. Table 1 provides the numerical values of zp for β = −5, −4.9, …, −0.1, 0.1, …, 4.9, 5, α = 1, λ = 1, μ = 1, θ = 0, and φ = 0. We hope that these values will be of use to the practitioners mentioned in Section 1. Similar tabulations could be easily derived for other values of α, β, λ, μ, θ, and φ. Sample programs are shown in the Appendix.

Percentage Points of Z

Acknowledgments

The authors would like to thank the editor for carefully reading the manuscript and for his great help in improving the article.

APPENDIX

The following programs in MAPLE can be used to generate tables similar to that presented in Section 3.

#this program gives percentiles when beta > 0

tt:=(mu/2)*exp(-mu*abs(y-phi)):

tt1:=(1/2)*exp(lambda*((z-beta*y)/alpha-theta)):

tt2:=1-(1/2)*exp(-lambda*((z-beta*y)/alpha-theta)):

ff:=int(tt*tt1,y=(z-alpha*theta)/beta..infinity):

ff:=ff+int(tt*tt2,y=-infinity..(z-alpha*theta)/beta):

p1:=fsolve(ff=0.6,z=-10000..10000):

p2:=fsolve(ff=0.7,z=-10000..10000):

p3:=fsolve(ff=0.8,z=-10000..10000):

p4:=fsolve(ff=0.90,z=-10000..10000):

p5:=fsolve(ff=0.95,z=-10000..10000):

p6:=fsolve(ff=0.99,z=-10000..10000):

print(beta,p1,p2,p3,p4,p5,p6);

#this program gives percentiles when beta < 0

tt:=(mu/2)*exp(-mu*abs(y-phi)):

tt1:=(1/2)*exp(lambda*((z-beta*y)/alpha-theta)):

tt2:=1-(1/2)*exp(-lambda*((z-beta*y)/alpha-theta)):

ff:=int(tt*tt2,y=(z-alpha*theta)/beta..infinity):

ff:=ff+int(tt*tt1,y=-infinity..(z-alpha*theta)/beta):

p1:=fsolve(ff=0.6,z=-10000..10000):

p2:=fsolve(ff=0.7,z=-10000..10000):

p3:=fsolve(ff=0.8,z=-10000..10000):

p4:=fsolve(ff=0.90,z=-10000..10000):

p5:=fsolve(ff=0.95,z=-10000..10000):

p6:=fsolve(ff=0.99,z=-10000..10000):

print(beta,p1,p2,p3,p4,p5,p6);

References

REFERENCES

Albert, J. (2002). Sums of uniformly distributed variables: A combinatorial approach. College Mathematical Journal 33: 201–206.Google Scholar
Ali, M.M. (1982). Distribution of linear combinations of exponential variates. Communications in Statistics—Theory and Methods 11: 1453–1463.Google Scholar
Chapman, D.G. (1950). Some two-sample tests. Annals of Mathematical Statistics 21: 601–606.Google Scholar
Christopeit, N. & Helmes, K. (1979). A convergence theorem for random linear combinations of independent normal random variables. Annals of Statistics 7: 795–800.Google Scholar
Davies, R.B. (1980). Algorithm AS 155: The distribution of a linear combination of chi-squared random variables. Applied Statistics 29: 323–333.Google Scholar
Dobson, A.J., Kulasmaa, K., & Scherer, J. (1991). Confidence intervals for weighted sums of Poisson parameters. Statistics in Medicine 10: 457–462.Google Scholar
Farebrother, R.W. (1984). Algorithm AS 204: The distribution of a positive linear combination of chi-squared random variables. Applied Statistics 33: 332–339.Google Scholar
Fisher, R.A. (1935). The fiducial argument in statistical inference. Annals of Eugenics 6: 391–398.Google Scholar
Hitezenko, P. (1998). A note on a distribution of weighted sums of iid Rayleigh random variables. Sankhya A 60: 171–175.Google Scholar
Hu, C.-Y. & Lin, G.D. (2001). An inequality for the weighted sums of pairwise i.i.d. generalized Rayleigh random variables. Journal of Statistical Planning and Inference 92: 1–5.Google Scholar
Kamgar-Parsi, B., Kamgar-Parsi, B., & Brosh, M. (1995). Distribution and moments of weighted sum of uniform random variables with applications in reducing Monte Carlo simulations. Journal of Statistical Computation and Simulation 52: 399–414.Google Scholar
Moschopoulos, P.G. (1985). The distribution of the sum of independent gamma random variables. Annals of the Institute of Statistical Mathematics 37: 541–544.Google Scholar
Pham, T.G. & Turkkan, N. (1994). Reliability of a standby system with beta-distributed component lives. IEEE Transactions on Reliability 43: 71–75.Google Scholar
Pham-Gia, T. & Turkkan, N. (1993). Bayesian analysis of the difference of two proportions. Communications in Statistics—Theory and Methods 22: 1755–1771.Google Scholar
Provost, S.B. (1989). On sums of independent gamma random variables. Statistics 20: 583–591.Google Scholar
Williamson, R.C. (1991). The law of large numbers for fuzzy variables under a general triangular norm extension principle. Fuzzy Sets and Systems 41: 55–81.Google Scholar
Witkovsky, V. (2001). Computing the distribution of a linear combination of inverted gamma variables. Kybernetika 37: 79–90.Google Scholar
Figure 0

Percentage Points of Z